Does Telemonitoring For Chronic Illness Work? The Emerging Body of Literature Says Yes.

That's the question that the Disease Management Care Blog asked itself after comparing this new study published in Health Affairs with this negative study that had been published in December 2010 in the New England Journal and reviewed by the DMCB.

Briefly, the new study, "Integrated Telehealth and Care Management Program for Medicare Beneficiaries with Chornic Disease Linked to Savings," involved patients with diabetes, heart failure or COPD cared for at the Wenatchee Valley Medical Center and the Bend Memorial Clinic in Washington and Oregon, respectively. 

It used “Health Buddy,” which is a handheld device with four buttons and a screen that is telephonically linked to a care management provider service.  The device asks a menu of condition-specific questions, and patient answers are aggregated by computer for subsequent review by the care management nurses.  If the answers are consistent with any deterioration, the patient is telephoned. 

Health Buddy was originally examined as a CMS Demo (the report is here).  This new Health Affairs study was a reanalysis of the data paid for by the owners of Health Buddy.

Briefly, in this new study, Medicare beneficiaries with diabetes, heart failure or COPD who appeared to be high risk and high cost were chosen by CMS.  Two batches of patients were chosen at different times: 763 were selected in early 2006 and, because of attrition and death, another 1056 were chose one year later.  37% of the candidates ultimately agreed to participate and use the Health Buddy device. There was no charge for using the device.  This was an "intent to treat" study, so even patients that declined participation were included in the analysis.

What was different in this analysis was that two control groups similar to the intervention groups were abstracted out of CMS' databases, using counties of residence elsewhere in the U.S. with similar degrees of urbanization, demographics and local care delivery systems. Claims and other data were then used in propensity matching to finalize a group of patients that resembled the intervention patients. 

A total of 52 patients were excluded because of problems with propensity matching, so the total analysis was based on 1767 intervention patients and 1767 (propensity-matched) controls.

Armed with a different control group, what did the authors find?

While death rates were slightly lower for the intervention group (10.5%) vs. control (10/7%)  in the first year, it was 2.5% lower in the second year (9.7% vs. 12.3%). 

At baseline,quarterly spending was $4048 for the intervention group vs. $4093 in the controls.  One year mean quarterly spending dropped to $3508 vs. $4107.  In the second year it was $3568 (intervention) vs. $4051 (controls).  Depending on which quarters were used, the authors estimated that total savings was $450 per quarter.  After the usual statistical adjustments to control for any observed variations and the ironic possibility that death may have result in savings to the Medicare program, the savings held up as statistically significant.  The statistical significance also persisted for each of the three disease groups; it was greatest for heart failure and lowest for diabetes.

How does the DMCB reconcile all this?

In contrast to the New England Journal study, where the results were forwarded to the patients' docs, Health Buddy had a nurse in the loop who called the patient at the first sign of any deterioration.  The DMCB believes that pairing dedicated staff to the telemonitoring made the difference; busy physicians are just not able to cope with additional information like this in typical day-to-day outpatient clinic workflows.

As mentioned above, this study was originally a CMS demo.  The original analysis of Health Buddy for CMS did not use propensity matching and, while costs went down, it did not achieve statistical significance.  The DMCB also thinks that the more analyses are conducted, the more disparate findings will emerge, but it also finds propensity matching to be a widely used and acceptable statistical approach.

All in all, the DMCB still thinks this is another study in an emerging body of science that supports the use of telemonitoring as one option in population health management and further evidence that disease management works.